Chinese entertainment for engineering students

“To sum up, if you want to achieve true independence, you must have your own instruction set.

Only you can provide all the support from software to supporting hardware, which is why I hope to establish a research program. "

"Are you planning to be the core?" Wang Xu listened with some surprise to the explanation of the thirty-year-old professor in front of him, and was quite surprised by their courage.

This project application meeting was arranged by Secretary Jiang. Now Wang Xu has opened the project application channels to the outside world. This is equivalent to providing the possibility of a relatively independent project application outside the school and the government. It is currently only open to internal people.

The applicant this time, Liu Weiping, comes from a numbered integrated circuit design center in the Imperial Capital. He has been deeply involved in the previous North and South Bridge projects and can be considered an insider. Obviously, their goal is not just the North and South Bridges.

Wang Xu participated in this project application meeting as the director of Wang Xu Laboratory. He is now fully responsible for the review of various new projects.

Today's projects are no longer limited to projects proposed by Wang Xu and then assigned by various departments. The opening of the application channel is to allow professors to use their brains more and enter the path of independent scientific research.

This can be said to be extremely rare in the 1990s. In a period of overall backwardness, it was what Wang Xu expected to be able to allow some of China's top scientists to break out of their ruts and think ahead about something other than catching up.

At this time, and even within the next ten years, domestic scientific research will be conducted in a way of writing to catch up and reading to copycat.

See what things are still not available in China, then study foreign products and gain something, that is victory.

Another method is to introduce, which means spending money to purchase foreign technology and then organizing domestic experts to digest and absorb it.

Most of these research directions are engineering, rarely involving principles and underlying science, but some practical benefits can be seen, and both the country and enterprises are willing to provide some funds.

A small amount of purely theoretical research is facing almost a comprehensive lack of funding and is struggling to make ends meet.

The channels provided by Wang Xu are not purely theoretical, but they are close to the bottom anyway, and they are basically things that cannot see any benefits in the short term. But the approval and review here will appear to be relatively simple and crude, that is, if Wang Xu sees the right person, he can vote.

"Yes, we want to do it. The final performance of your supercomputing is quite outstanding. It has far exceeded the level of foreign 486 workstations. It shows that the idea of ​​​​your project itself is very good.

But obviously, you lack cores. The reason why you didn't use 486 is that co-processors are not easy to buy, so you settled for the next best thing. But we can definitely build an independent co-processor, one that specializes in floating-point operations.

You once mentioned in supercomputing design that a large number of identical small cores can be integrated into one unit to perform parallel operations to replace the existing multi-board model. I think this is definitely a new type of chip and has absolutely wide uses. Very.

After the completion of the last project, we set up a team and expanded it according to your ideas. The content just explained is our latest achievement. I hope Professor Wang can seriously consider it.

It is better to have one yourself than your father has. This kind of chip is not yet available abroad. We can start with a few cores and then expand the number of cores from generation to generation. This is definitely the best starting point for researching new instruction sets.

Streamline instructions, parallelize, do not need internal cache, unify external cache, multiple pipelines, high pipeline, and multiple channels.

This is the novel model you proposed. We can completely streamline it into one unit, and still support the multi-board model. In this way, the number of cores equivalent to M times N can be operated at the same time. It feels wonderful to think about it. It can definitely be an absolute powerful tool for supercomputing in parallel operations.

Our individual films are weaker, but they can be stacked in this way to continue to expand your basic concept of supercomputing and throw a lot of computing work to these computing units to complete.

The host part can greatly reduce the computing requirements. It only needs to complete task distribution and human-computer interaction. Isn't this the real future supercomputing you mentioned in your plan?

Look, I heard that you are planning to set up a second-generation workstation project, but the name is supercomputer. Isn’t it just a false name to build a workstation? Why not really make it a supercomputer? "Liu Weiping's words were like a storm, sonorous and powerful. He was like this when he was explaining just now, and he seemed particularly powerful.

This person thinks quickly and speaks very fast. If you don't have any skills, you can't even keep up with his speaking speed. Wang Xu didn't even have time to reply. He talked a lot, from basic technology to future development, which is called comprehensive.

Well, Wang Xu actually had some impression of this person, and he had dealt with him in his previous life. He later became one of the core leaders and a senior executive of BGI Jiutian. But their main product is EDA, which is software.

It was only this time that I learned that he was also engaged in integrated circuit core research and development at the beginning. He was also an alumnus and had a PhD in computer science after graduating a few years ago.

The biggest advantage of top universities is the almost ubiquitous connections. If you don't pay attention, you will find an alumnus hiding somewhere. This is the most important reason why Wang Xu has been staying in Shuimu.

It was Mizuki's resources that allowed him to get through the novice stage almost without any obstacles, grow into a boss in a very short period of time, and now he has the right to approve projects.

"Your introduction is fine. There are some details in the idea that need to be discussed. But the general direction is correct and I agree with it. But do you only work on the co-processor? What about the main core? What if the instruction sets are different? How to coordinate?”

Wang Xu's problem is very acute. If this problem cannot be solved, it is meaningless to study a coprocessor alone to perform floating point operations.

"This is also the main reason why we came to you. Our opinion is the same as last time, let's do it together.

It's still like the North and South Bridge project, which involves multiple agencies working together. The cooperation last time was very pleasant, but the premise is that you provide a fairly detailed underlying design here.

I hope you can take the lead this time. If the underlying design remains consistent, both the central processor and the co-processor can be designed together.

Moreover, the several equipment groups and software groups that were combined last time should also be introduced together. The new instruction set also needs the support of the operating system. This is another big project, and you still have to come forward to unify it.

In this way, the second-generation supercomputer can become a true supercomputer. The floating-point calculation speed of the first-generation supercomputer has exceeded 100 million times, fully reaching the level of the Galaxy supercomputer ten years ago.

But this speed is not enough. Internationally, it has reached one billion times. But if our next-generation supercomputer is implemented in this way, 10 billion times should not be a problem. "Liu Weiping really talks a lot. Wang Xu rarely encounters people who talk more than himself, and he saw it today.

However, what he said is reasonable. The last project has been basically completed, and the development of the next generation product has indeed started. This thing inherently requires constant iteration.

"It's okay to think about it this way, but this project is huge. I have to think about it comprehensively. Just wait for my news. I will give you a reply within a week at the latest." Wang Xu doesn't plan to continue chatting. There is no problem with this project.

In fact, Wang Xu had already decided to start the research and development of the second generation of "supercomputer", even without Liu Weiping's trip.

However, it is better if the project is proposed by them, which lays the foundation for long-term cooperation in the future. Fully mobilizing the enthusiasm of all participating units is not only about money, but also about emotions.

They are not from the same unit. The future is more about cooperation, and the relationship between them still needs to be good. They proposed and adopted it here. This is all a favor, and it is not the same as a completely unilateral assignment of tasks.

The Tongfang workstation is already applying for national acceptance. This is a big project. In the end, there must be a reasonable summary, a series of machines are laid out, and various test indicators are listed.

Although it cannot be evaluated in various ways like later generations of networks, it is still necessary to compare it with workstations such as SGI and Sparc on the market.

Wang Xu is not worried about this aspect. He has already used this set of things. Although there are still some bugs, it is generally very advanced. Except for the 19-inch screen, which looks too small, and the resolution of 1280X1024. Not very high.

But it is more than enough for audio synthesis, and Wang Xu is not greedy.

To put it bluntly, this time it is about team building. Many projects are even developed at the bottom level. It is unexpected to be able to produce so many results, okay?

The future is the time for the development of various applications. He plans to list the projects and then slowly consider priorities.

Even though it has not been accepted yet, the computer company Shuimu Tongfang is already working on the commercialization of the workstation, and will authorize several companies to work together in the future.

Fangzheng from Imperial University and Inspur from Qilu have already signed up, and they are expected to get the license soon. The country will not allow such an important thing to be monopolized by one company.

Founder first started his business by making laser phototypes, and he can be said to be the leader in the field of printing. Recently he also started making computers, but his sales were not very good. Shuimu Tongfang was established much later, and he was basically established in this life.

Even the name has the meaning of being different from Fang Zheng.

Although Inspur also makes personal computers, it is not its main business scope. They mainly target the server market.

They just launched their own minicomputer server product this year. In terms of computing performance, it is not as good as their own.

However, the network processing capabilities of the server are first-class, so they plan to combine the two technologies to build a new architecture server.

Tongfang is mainly dealing with the field of scientific computing. The current top-level single device has over 100 million floating-point operations. This is already at the level of current minicomputers, and its performance is quite excellent.

The current benchmark company and product is Chouguo's SGI workstation, which is the kind of workstation used to make the liquid robot in "Terminator".

In fact, there are not many real computer special effects shots, most of them are special effects.

For example, in the scene where the T800 sank into the molten steel at the end, milk was used underneath, and then a strong orange light was used to create the effect of the molten steel. Therefore, Arnold just took a milk bath in the end, which was quite beautiful.

Only the shot of the liquid robot standing up from the ground was a real computer special effect.

Therefore, even now, the proportion of computer special effects in movies is still very small, and the main special effects models, including those in Hollywood, are still based on traditional methods.

Now that Wang Xu is in the game, there won't be many technical shortcomings at all. Currently, dedicated 3D processing software products are already available in the open source community, but they have not been fully developed yet.

Moreover, the technical foundation of this software is a bunch of scientific computing content, including physics engines such as ray tracing, gas simulation, and explosion simulation, which are all compatible with that part of scientific computing software.

Therefore, Wang Xu took out the core function packages of the basic computing part separately and existed as independent open source projects.

When these scientific computing things are completed and a suitable interface is added, it will become the famous MatLab in future generations. However, MatLab is a commercial software and is not open source, while Mizuki's set is open source.

Although no one has the energy to develop packaging packages yet, I believe that within a few years, this thing will have an impact on the MatLab market.

The final 3D engine focuses more on interaction and the connection between various physics engines.

This engine is highly connected with 3D CAD software and 3D animation software in many places, so it has been divided into small open source projects for development as public parts.

The splitting and combining of these software engineering aspects is what consumes the most energy of Wang Xu, and is also his most important job as the project leader.

At present, the projects in the open source community are still very limited, but many excellent projects have been born. For example, the audio processing software that Wang Xu has been thinking about has been successfully developed.

It is enough to meet the needs of student Wang Xu for graphical drag and drop capabilities of various audio synthesis.

At that time, in order to realize this function, Wang Xu also developed a graphical interface. What about the style? Classmate, have you heard of Windows 98?

Anyway, the essential core of graphical interaction is actually the various underlying interfaces. The surface interactive pages can be designed according to preferences to develop different interface styles, Windows 98 style.

It was caused by Wang Xu's bad taste. The main theme is to follow other people's paths and leave others with no way to go.

The graphical interface itself was developed in C++. Most of the main engineers came from the Institute of Computing Technology of the Chinese Academy of Sciences. This part was taken over by them. They are a group of top-notch programmers, plus a very mature underlying framework design given by Wang Xu. , making the overall operating effect quite explosive.

Even on the most basic version of a single-CPU workstation, this set of graphical interfaces can run very smoothly and stably. Of course, some minor problems are inevitable. It is still being tested by the open source community from all over the world.

Wang Xu even provided a language dedicated to developing graphical interfaces early, named C # , borrowing the name from later Microsoft. The syntax is similar, but the class library is completely different.

The C # given by Wang Xu can be almost perfectly integrated with the C language. With the help of development tools, it can become a layer of skin wrapping the C language, which greatly simplifies the development process of graphical interactive functions.

And because this set of C # memory management is based on pre-allocated virtual memory, there is almost no memory overflow problem, so the final stability of the product is easier to ensure. This is specially prepared for ordinary programmers. A powerful tool for interface development.

Moreover, this set of things has also been borrowed into the development process of the 3D engine, completely separating the graphical display and complex background calculations, making it possible to achieve simple development and high operating efficiency.

This is one of the biggest benefits of having a unified planner for the product. There are so many places that can be reused. Because of this, the various standard libraries that Wang Xu has created here are all experienced hands. It can be said to be used almost everywhere.

——

(See you tomorrow. Chapter 4 has been revised twice and I haven’t passed it yet.

In this chapter, I accidentally brought some chip stuff with me. With the blessing of Sanqing, I passed the trial in one go.

It's harder to change than writing about Sese. The key is that you don't know what is banned. )